我的解决方案是将responses parts、responses和swagger_auto_schema分别放在3个不同的地方。在我的项目根目录下，我创建了utils文件夹，在其中我以如下方式存储文件：

utils
├── __init__.py
├── ...
└── swagger
   ├── __init__.py
   ├── generators.py            <---- http/https schema for swagger
   ├── response_templates.py    <---- templates of swagger_auto_schema response parts
   └── serializers.py           <---- some general response serializers

在response_templates.py中有以下代码：

from drf_yasg import openapi
from rest_framework import status
from .serializers import DetailResponseSerializer, TaskRequestSerializer
FILE_SUCCESS_200 = {
    status.HTTP_200_OK: openapi.Response(
        description="Correct file get",
        schema=openapi.Schema(type=openapi.TYPE_FILE),
    ),
}
TASK_CREATED_201 = {
    status.HTTP_201_CREATED: openapi.Response(
        description="Successfully task created",
        schema=TaskRequestSerializer,
    ),
}
... # and other

因此，我可以在我的swagger_auto_schema装饰器中使用这些响应部分。下一步是在所有应用程序中创建responses文件夹，那些需要在swagger.tree of responses文件夹中自定义响应的应用程序：

responses
├── __init__.py
├── image.py
└── and_other_models_name_files.py

在image.py中有以下代码：

from utils.swagger.response_templates import (
    TASK_CREATED_201, FILE_SUCCESS_200, CREATED_201, SERIALIZE_400, NOT_FOUND_404,
    LARGE_ENTITY_413, INVALID_FILE_FORMAT_415, SERVER_STORAGE_500
)
image_list_responses = {
  **TASK_CREATED_201,
  **SERIALIZE_400,
  **LARGE_ENTITY_413,
  **INVALID_FILE_FORMAT_415,
}
image_detail_responses = {
  **FILE_SUCCESS_200,
  **NOT_FOUND_404,
  **SERVER_STORAGE_500
}

因此，作为上述重构的结果，我可以在swagger_auto_schema装饰器中使用上述响应。

from media_app.responses import image_detail_responses
class DetailDownloadAPIView(RetrieveDestroyAPIView):
    @swagger_auto_schema(responses=image_detail_responses)
    def get(self, request, *args,**kwargs):
        ...
...